The aim of this in vivo study was to radiographically evaluate the proximal contour of composite resin restorations performed using different matrix systems. Patients with premolars needing class II type resin composite restorations involving the marginal ridge were selected. Thirty premolars were selected and randomly divided into three groups (n = 10 each) to receive restorations using different matrix systems: group 1: metal matrix coupled to a carrier matrix and wood wedge (G1-MMW); group 2: sectioned and precontoured metal matrix and elastic wedge (G2-SME); and group 3: a polyester strip and reflective wedge (G3-PMR). After the restorative procedure, bitewing radiographs were performed and analyzed by three calibrated professionals. The quality of the proximal contact and marginal adaptation of the proximal surfaces was classified as either correct or incorrect (undercontour/overcontour). The Pearson Chi-square statistical test (α = 5%) revealed a statistically difference between frequencies of correct and incorrect restorations (α(2) = 6.787, p < 0.05). The group G2 SME produced a higher frequency of correct proximal contours (90%), while G1-MMW and G3-PMR had a ratio of 40% correct and 60% incorrect contours respectively. None of the matrix systems was able to prevent the formation of incorrect proximal contours; however, the sectioned and precontoured metal matrix/elastic wedge configuration provided better results as compared to the other groups.